Candidate password generation and application cracking method on GPU

Abstract

The invention discloses a candidate password generation and application cracking method on a GPU. The password generation method comprises the following steps: 1) creating a cracking task according toa target demand, and setting the size N1 of a cracking space; 2) processing a training set to obtain context-independent grammar and rule probability; 3) calculating a maximum candidate password number N2 which can be generated according to the context-independent grammar, and determining a total candidate password number N = MIN (N1, N2) generated by the cracking task; determining the number ofcandidate passwords generated by the kernel function each time, the total number of started threads and a memory space on the GPU required for running the kernel function according to N and GPU performance; 4) allowing the kernel function to generate candidate passwords according to the context-independent grammar and the number of the candidate passwords generated by the kernel function, determining a basic grammar structure in a basic grammar rule table processed by the kernel function, and generating the candidate passwords meeting grammar requirements by utilizing the GPU; and 5) returningto the step 4 until all candidate passwords are generated.

Description

technical field [0001] The invention relates to the technical field of password cracking, in particular to a PCFG model-based candidate password generation and application cracking method on a GPU. Background technique [0002] Currently, well-known password cracking tools include Hashcat and John The Ripper. Among them, John The Ripper supports many types of cracking algorithms and various cracking modes, but it mainly uses CPU as the computing unit, and does not support GPU parallel acceleration very well. Hashcat uses the Opencl programming language, supports multiple platforms, and has a complete range of algorithms, especially for GPU parallel acceleration and optimization. It is known as the fastest and most advanced password cracking tool in the world, so we take Hashcat as an example. It mainly provides the following cracking modes: violent mode, dictionary mode, dictionary + rules and other modes. [0003] Among them, the brute force mode is also called exhaustive ...

AI Technical Summary

Problems solved by technology

[0006] The existing attack mode that can support PCFG basically separates the cracking process from the process of using the PCFG model to generate a password dictionary, first generates a password dictionary through the PCFG model, stores it in the local hard disk, and then reads the generated password dictionary to The CPU is copied to the GPU memory, and finally the password in the GPU memory is involved in the cracking process. This process involves dictionary generation, storage, reading, copying, etc., and is not a real GPU-based PCFG attack mode.

[0007] To sum up, we can find the existing problems in candidate password generation in real cracking scenarios: 1) Candidate passwords generated through brute force mode pay less attention to the sequence of passwords and the internal correlation of generated password characters, and need to try a large number of low-quality passwords. When the real password is long, it needs to try a huge number of candidate spaces, which makes the cracking efficiency low; 2) The candidate password generated through the dictionary mode is closer to the real password, but involves the process of generating, storing, and reading a large number of dictionaries , further because of the transmission bandwidth between the CPU and the GPU, loading it into the GPU memory will seriously restrict the overall algorithm cracking speed

[0008] The present invention provides a PCFG model-based candidate password generation and application cracking method on the GPU. The method can support the GPU-based PCFG attack mode in a real cracking scene, and expand a large-scale candidate password that is closer to reality through a small dictionary. Password, while avoiding the problem that the transmission of massive dictionary data from the CPU end to the GPU end limits the overall cracking speed

Images